Integration of 10X genomics single-nuclei RNA seq datasets

Question

[paria]

I have the single nuclei datasets (10X genomics) from 4 different regions of brain and spinal cord of two groups (patients and control). I am trying to compare different regions within patients (resistant and vulnerable region to pathology) and between patients and controls. Actually, I have different comparisons. I am wondering should I integrate the pair I am going to compare or I should integrate all them together then do the comparison. I think that first one should work better but I need to know what do you recommend? and if I integrate all of them together I think the clustering would be different from the first approach. I'm kind of confused about that. Any help would really appreciated. Thanks, Paria